On Fri, 2008-11-14 at 09:00 -0600, Dan Armbrust wrote:
> > There was concurrent access to the table during VACUUMing, so the long
> > delay is explainable as long waits for cleanup lock, plus probably
> > thrashing the cache with bloated indexes. The CPU overhead per row seems
> > OK. We should i
>
> There was concurrent access to the table during VACUUMing, so the long
> delay is explainable as long waits for cleanup lock, plus probably
> thrashing the cache with bloated indexes. The CPU overhead per row seems
> OK. We should instrument the wait time during a VACUUM and report that
> also.
On Thu, 2008-11-13 at 17:30 -0700, Scott Marlowe wrote:
> > I had them run a vacuum analyze verbose on my database, and had these
> > lines come back which made me suspicious:
> >
> > INFO: index "ix_cpe_ispid" now contains 41626 row versions in 13727 pages
> > DETAIL: 5224 index row versions w
Thanks everyone. You have helped back up my suspicions. It is indeed
a Linux system, and it has a RAID IO system, but I don't yet know the
details of that IO system. Time to put them back to work looking at
their hardware, rather than blaming our software :)
Thanks for the extra tips on hunting
Scott Marlowe wrote:
On Thu, Nov 13, 2008 at 4:08 PM, Dan Armbrust
<[EMAIL PROTECTED]> wrote:
Why did those particular tables and indexes take _so_ long to vacuum?
Perhaps we have a disk level IO problem on this system?
Assuming pagesize is 8k, then we're talking about scanning 1303*8192
byt
"Dan Armbrust" <[EMAIL PROTECTED]> writes:
> Why did those particular tables and indexes take _so_ long to vacuum?
> Perhaps we have a disk level IO problem on this system?
FWIW, I agree with Scott that you seem to have an overstressed I/O
system. It's hard to tell why from here.
> Can someone t
On Thu, Nov 13, 2008 at 4:08 PM, Dan Armbrust
<[EMAIL PROTECTED]> wrote:
> I have a system backed by a PostgreSQL DB at a customer site that
> mysteriously slowed way down - and couldn't keep up with the load for
> no apparent reason.
>
> I had them run a vacuum analyze verbose on my database, and
